Man I’m ready to build a new system, spare no expense but I’ve just plan had it!!!
All I wanted to do is play some BF6 n other games but think it’s over, let WWIII commence…..

If it's something i learned in the more than 20 years i had this hobby, is that you never wait if you have the spare cash to buy hardware. Software always goes down in price, but hardware always was up and down for all sort of reasons, natural disasters, war, global events or corporate greed.
What i am glad, is that i kept my older hardware and upgraded it so that i have a backup PC and a main one.

droopyROIf it's something i learned in the more than 20 years i had this hobby, is that you never wait if you have the spare cash to buy hardware. Software always goes down in price, but hardware always was up and down for all sort of reasons,

Eh? For the latest and greatest, software goes up in price faster than hardware. The Mathematica package I first bought for $400 cost me $4K last year. Autocad subscriptions have doubled in price in the last decade.

If you're talking performance-constant dollars, then over the last 30 years, anyone who randomly purchased hardware in advance by a year or more on average wasted most of their money. Price hikes tend to be rare, and short-lived.

YouseethisFrom my little corner of the world: small to intermediate size suppliers (they supply stores/small store chains) have indeed, low stock on the 5060 but the price hasn’t approached the 5060 ti 8gb (yet?). The 5080 on the other hand increased in price since the beginning of the week. It holding up until now at least the cheapest, most unwanted models. But now they are almost gone, burning through the last cheap remaining stock. The next batch is going to be more expensive (by 10-20% maybe? I will find out next month)

The next batch is here sooner and it is indeed 10-20% more expensive.

EndymioEh? For the latest and greatest, software goes up in price faster than hardware. The Mathematica package I first bought for $400 cost me $4K last year. Autocad subscriptions have doubled in price in the last decade.

If you're talking performance-constant dollars, then over the last 30 years, anyone who randomly purchased hardware in advance by a year or more on average wasted most of their money. Price hikes tend to be rare, and short-lived.

To clarify about software, i was talking about what a home user uses, like games and such. If you use the above mentioned you have a business and it is an investment.
I would not call 2-3 years "a short period", the two big crypto booms we had in the 2010's and the one in 2020-2023.

But those days are gone, that is my point, AI is here and never going away, it is not a bubble and nothing will be the same as it was in 2025 or early years.

droopyROTo clarify about software, i was talking about what a home user uses, like games and such. If you use the above mentioned you have a business and it is an investment.

Games have escalated in price by similar margins. I remember when a game over $10 was expensive ... today, you can pay well over $100 for many "ultimate edition" AAA titles, no?

droopyROI would not call 2-3 years a shor period, the two big crypto booms we had in the 2010's and since 2020-2023.

3 years? The last video card price hike lasted from Nov 2020 to Jun 2022 -- 18 months. But there was only one 6-7 month period during that time for which you'd have saved money by purchasing a GPU a year plus in advance. And while I didn't check price history for other components, I don't think was any such period for anything but GPUs.

droopyROBut those days are gone, that is my point, AI is here and never going away, it is not a bubble and nothing will be the same as it was in 2025 or early years.

The laws of economics haven't changed. AI is currently financing a massive Capex expansion in the industry. Sometime in mid 2027 (my estimate) or early 2028 (most other estimates), consumers purchasing consumer-grade GPU and RAM will reap the benefits of that.

Back when in the '80s ? You can find good prices in Steam/Epic/GOG sales, there are at least 3 each year, there are a lot of sites like Instant Gaming, Fanatical, Kinguin etc. that have offers and free games weekly like the ones on Epic. Also there is Humble Bundle and Amazon with monthly games packs with great value.

And that is my point, PC games have never went up in prices. Never has a game that launched at 60$ cost 80$ after 2 years as it happened to GPU's in the crypto era or SSD/RAM/storage today.

18 month or 24 dose not matter, it is a huge amount of time for any human being. GPU's cause that is what matters here on this forum, being a hardware enthusiast/gamers oriented site.

The benefits of what ? What is a benefit of AI, i can't think of one. And please don't tell me to cure cancer or some other BS. None of us will ever benefit from that.

droopyROBut those days are gone, that is my point, AI is here and never going away, it is not a bubble and nothing will be the same as it was in 2025 or early years.

While it is not going away, it is bubble for sure.

And it is huge bubble, one that i neccessary to keep printing machine, and one that is doomed to fail.

Simply, running costa of general AI are too high for it to be freely available... And they are too high due to vasic laws of physics, so it is not going to change soon.

However, as a highly specialized and paid tool, AI can and will be awesome. But you cant keep printing machine running on such tool, no?

Looking at the bottom of this page, it shows threads that are related to this thread. If you just read them straight down the line it seems this is all just fear mongering. Like a cool way to artificially make folks scared into making 'last one' purchases. Social manipulation probably constructed by Mr. AI Bundy. That is my new name for it since a capitol (i) is the same as a lower case (L)

I smell rotting flesh.

droopyROI agree. What @Vayra86 is missing, is that this time is not only individuals and some people with cash that want quick money.

With AI we are talking all branches of governments, all businesses from mom and pop stores and individuals that work at home to trillion $ corporations, all use and will use AI.
And the cherry on the cake is the militaries of the world, cyberwarfare, electronic warfare, drone warfare and any other type of sea, land, air or space warfare will have AI integrated in some way. The AI era is just starting ...

Im not missing that. Still there is major overspending with questionable ROI. It aint gonna last. Where are all these massive leaps we are making because of AI? All we hear and see is the money spent on it and hallucinations and pretty weak creative content.

Those datacenters certainly arent needed for military or gov purposes, they dont need that kind of volume and if they did, again... that would make their new MO far too vulnerable and costly.

Shit just doesnt check out and something is gonna give.

Hecate91I wish I could be so optimistic I really do, but no this is 100X worse than crypto.
During the crypto scam I could buy every other part except a GPU, or find used deals on a GPU, now thanks to the AI slop you can't buy anything, or you have to pay up the price gouged prices or just have an empty PC case to look at especially for anyone who missed out on normal priced hardware.
Every major corpo is in on the AI bubble, governments are handing out piles of money to AI companies so we'll be forced to use AI one way or another, as a matter of control of what you can do on your PC, if PC's and parts will even be obtainable anytime soon. Several tech CEO's have admitted they don't want the average user to have local computing power because that means we won't have to rely on renting computing power in the cloud. None of these AI companies or those involved except Nvidia are making a profit, CEO's are getting desperate telling people that AI is going to be the best thing ever, but the desperation and no ROI doesn't seem to matter when investors keep getting gaslit over AI being anything more than a bunch of chatbots and LLM's consuming massive amounts of electricity and water.

In the end you cant sell air, and AI for the vast majority of use cases simply is hot air. Inefficient, and gloriously blunt and lacking at its core. The quality of LLMs is already reaching a plateau and the noise around AGI is slowly dying down and is on the same trajectory as VR, AR, Crypto as the free man's coin and various other tech bro fantasies. The current state of it all certainly isnt good enough to make bank on all those promises it holds. Not even remotely close.

It aint worse. Its just another thing that will pass.

droopyROTo clarify about software, i was talking about what a home user uses, like games and such. If you use the above mentioned you have a business and it is an investment.
I would not call 2-3 years "a short period", the two big crypto booms we had in the 2010's and the one in 2020-2023.

But those days are gone, that is my point, AI is here and never going away, it is not a bubble and nothing will be the same as it was in 2025 or early years.

Crypto isn't gone either... and yet, its not gobbling up the GPU market anymore.

blog.kraken.com/crypto-education/crypto-markets-in-2026

You know what all of this is? Shitty nobodies trying to make bank on some nonachieving bullshit, very much akin to the average political discourse in several countries not accidentally the ones that are biggest and all-in on AI and crypto ;). Interesting trend I would say, as it reveals the reality: the 1% trying to exercise force to pull even more assets into their garden, trying to control the flow of information to get there even and execute their pyramid scheme. The industry invests in itself and cheers how much money's going around in it, that is what AI and Crypto at its core really are, and they hope to reach some level where nobody can get around them.

But... Where is the critical mass in the userbase? The people themselves? Like intrinsically, the way we use a search engine, that kind of critical mass. It isn't there. Its being marketed as if it should be there, we get free tools, and we get to discover all the shit we can pull with it. Tech thinks this will work just like how Google conquered us with their search engine. Look how they integrated it in that same thing, even. It is working for you whether you want to or not. But the irony here is: that search engine spits out better results without AI. Anyone taking a bit more effort than reading the summarized AI slop, will end up knowing more. That's really all you need to know, to realize this shit is dead in the water. The majority just hasn't figured that out yet, but they will. Now sure you could build a better tool to assist in search. But that takes massive effort and specialization, which is fighting economic feasability. This technology, like crypto, is caught in a catch 22, because the previous solution was already good enough.

If you ask me, crypto had a much more dedicated core of actual community push than AI.

Vayra86But... Where is the critical mass in the userbase? The people themselves? Like intrinsically, the way we use a search engine, that kind of critical mass. It isn't there. Its being marketed as if it should be there, we get free tools, and we get to discover all the shit we can pull with it. Tech thinks this will work just like how Google conquered us with their search engine. Look how they integrated it in that same thing, even. It is working for you whether you want to or not. But the irony here is: that search engine spits out better results without AI.

Actually, since AI summaries, Google search went to hell. Bing also... They are leaning so hard to AI that you actually get some information through AI (that then you need to check manually to be sure), and complete bull***t via Search.

But, theoretically you could push huge number of users to use AI daily, just like they use Search. That is why we have "Copilot New Page", "Comet AI browser", Google AI summaries, so they can point out millions of people use AI daily. So, they will prove millions of users to investors, no problem.

Issue is - Google Search, in original, was very efficient system. Modern-day AI is not, not at all. Huge resources are needed for daily operation, and they can not be offered for free to end-users. Yet, everybody are offering it for free, burning investors money and hoarding huge amount of user data that will, they hope, allow them to sell anything and "convince" users of anything. That alone would be enough, but I think that train sailed long ago... Geopolitics is not the same as in early 2000s, and sooner or later everyone in this game will have to admit AI is not "miracle" that will save tons of money, it is just another tool - good tool, but not magic bullet for everything.

Vayra86In the end you cant sell air, and AI for the vast majority of use cases simply is hot air. Inefficient, and gloriously blunt and lacking at its core. The quality of LLMs is already reaching a plateau

The disinformation of the above statement isn't just the hyperbole of the first sentence, but that, once again, someone who knows nothing whatsoever about machine learning is equating AI with LLMs. LLMs aren't even the entirety of generative AI, without even considering predictive AI.

Furthermore, the claim that LLMs are "plateauing" is about as far from truth as possible:

magazine.sebastianraschka.com/p/state-of-llms-2025

Nearly every major firm in the world is implementing some form of AI at present. And in just the last 2 years, the share of those firms which are *already* receiving ROI on their AI investment has risen from just 10% to nearly 75%.

knowledge.wharton.upenn.edu/special-report/2025-ai-adoption-report/

These are facts. The anti-AI crowd may not like them, but they need to deal with them.

The disinformation of the above statement isn't just the hyperbole of the first sentence, but that, once again, someone who knows nothing whatsoever about machine learning is equating AI with LLMs. LLMs aren't even the entirety of generative AI, without even considering predictive AI.

EndymioFurthermore, the claim that LLMs are "plateauing" is about as far from truth as possible:

magazine.sebastianraschka.com/p/state-of-llms-2025

Well, if you carefully read the article, you will see that plateau is actually real.

EndymioNearly every major firm in the world is implementing some form of AI at present. And in just the last 2 years, the share of those firms which are *already* receiving ROI on their AI investment has risen from just 10% to nearly 75%.

knowledge.wharton.upenn.edu/special-report/2025-ai-adoption-report/

Tech-bros babble, like I am reading one of their famous NFT reports.

EndymioThese are facts. The anti-AI crowd may not like them, but they need to deal with them.

No-one said that AI have no future - general AI do not have future because they are too expensive to be used as replacement for "everything" on the internet. Highly-specialized and trained AI is tool of future.

Keyword: TOOL.

Not a magic bullet for gathering zillions of USD of investments because it will revolutionize... anything.

It is evolution. Good one. But not something that will suddenly transform Earth to techno-paradise. This is not 2015 anymore, people are getting tired of all technological revolutions being fantasized about, while economy is going down to toilet everywhere. Or you still believe in Musk promises?

Aleksandar_038Well, if you carefully read the article, you will see that plateau is actually real.

You mean, when it states something like this, in the very first paragraph:

"...2025 was a very eventful year for LLMs and AI, and this year, there was no sign of progress saturating or slowing down."

Don't misrepresent articles to support your beliefs.

Aleksandar_038Tech-bros babble, like I am reading one of their famous NFT reports.

Nonsense. The report was compiled by professors at U Penn's Wharton School of Business, the #1 ranked business school in the world, and is based off real-world data from surveying 800+ corporations with active AI implementation projects.

Aleksandar_038No-one said that AI have no future - general AI do not have future because they are too expensive

Unsure whether by "general AI", you actually mean generative AI, or artificial general intelligence. In the first case, you're quite wrong, and the second case remains to be seen. Certainly AGI is a bit further away at present than many journalists are predicting. My own belief is that AGI won't arrive until neuromorphic hardware replaces the current GPU-driven approach.

Aleksandar_038Keyword: TOOL.

Not a magic bullet for gathering zillions of USD of investments because it will revolutionize... anything.

I'm hard-pressed to name one major industry that AI won't revolutionize. Care to help us out?

EndymioYou mean, when it states something like this, in the very first paragraph:

"...2025 was a very eventful year for LLMs and AI, and this year, there was no sign of progress saturating or slowing down."

Don't misrepresent articles to support your beliefs.

Well, like I said I suggest you read full article.

There is no contest that 2025 was eventful year and that 2026 will be even more eventful. However, if you read whole text instead of only first paragraph, it also states issues that LLM are running into lately, as well as range of innovations to overcome those issues. However, it is all very much WIP, and hardly turnkey solution - yet tech bros, Altmann and jacketman convince us it is all green in paradise.

Article also states that current business model is hardly sustainable in long run...

And that is exactly what others said here - business model is unsustainable. If you pay 30$ per month for some amount of usage, and that is only way to recoup your electricity and hardware costs, for how long you will survive on the market?

Nobody claimed that AI is useless. We just contest simple fact: tech bros told everybody that they will replace expensive programmers with AI, and in real life, we are watching free fall of software quality due to vibe coding 😁

Even article confirms that AI can be great tool for programmer, but never replacement - yet, we are constantly told different story by investors and high ranking industry officers...

I said before - AI will be great tool, but only when used as highly specialized tool. As free magic bullet for all, it is only money laundering machine.

EndymioNonsense. The report was compiled by professors at U Penn's Wharton School of Business, the #1 ranked business school in the world, and is based off real-world data from surveying 800+ corporations with active AI implementation projects.

Well, business school is hardly what I call reliable source. They can claim to be best they are still just bunch of salesman trying to sell you something every time when they open mouth.

That "best business school in the world" even had several articles about huge NFT potential, when everybody and everyone clearly saw what idiocy it is. That is about as low as one can get.

So, I like to listen serious programmers, engineers, chemists, physicst, hell economists... But bunch of salesman - no, thank you.

P. S. And for one serious report, it is missing quite a lot of references 😊

EndymioUnsure whether by "general AI", you actually mean generative AI, or artificial general intelligence. In the first case, you're quite wrong, and the second case remains to be seen. Certainly AGI is a bit further away at present than many journalists are predicting. My own belief is that AGI won't arrive until neuromorphic hardware replaces the current GPU-driven approach.

Generative AI, and yes - under current level of technology, it is unsustainable. It opposes basic laws of physics, therefore it is doomed to fail - or change and evolve if it wants to survive.

AGI is pipe dream, we have decades of hardware development in front of us before we can think about it.

EndymioI'm hard-pressed to name one major industry that AI won't revolutionize. Care to help us out?

Well, development of ironworking revolutionized war. Development of gunpowder, too. Development of antibiotics revolutionized medicine.

Development of internet also revolutionized all industries. So, AI will continue in same direction, I dont contest it.

I contest usage of AI as general purpose tool for everything - it is sustainable to use (and pay a lot of money) for AI that will increase your productivity. But to do same for free, or to use it as a search tool, given running costs, it is not sustainable at all.

Practical example: lawyer can speed up his work greatly, radiologist as well. But as long as they take responsibility for their work, they are irreplaceable and story sold us by tech bros is just money laundering machine.

Of course, if AI owner is willing to take responsibility for its mistakes, then maybe lawyer or radiologist can be replaced, but I am quite sure that we are decades from that scenario.

Or, you can overload one lawyer with three times more work because he has AI to help him - however, quality of work will go down. Just look at all vibe coding achievements recently...

Is this for mobile AND the desktop form?

EndymioThe disinformation of the above statement isn't just the hyperbole of the first sentence, but that, once again, someone who knows nothing whatsoever about machine learning is equating AI with LLMs. LLMs aren't even the entirety of generative AI, without even considering predictive AI.

Furthermore, the claim that LLMs are "plateauing" is about as far from truth as possible:

magazine.sebastianraschka.com/p/state-of-llms-2025

Nearly every major firm in the world is implementing some form of AI at present. And in just the last 2 years, the share of those firms which are *already* receiving ROI on their AI investment has risen from just 10% to nearly 75%.

knowledge.wharton.upenn.edu/special-report/2025-ai-adoption-report/

These are facts. The anti-AI crowd may not like them, but they need to deal with them.

LOL these shitty sites are the exact same kind of sources that get produced by the crypto community. The similarities are unreal.

Agentic AI for enterprise is where its at, but that is not an 'AI for most users' at all. For 'most users' its brutal overkill and totally unnecessary.
I work for a global IT firm, I'm pretty close to the actual material in actual companies. The most valuable Agentic AI is the one that has a playground made for it, and strict guardrails and monitoring / adjustment.

Vayra86LOL these shitty sites are the exact same kind of sources that get produced by the crypto community. The similarities are unreal.

The website of the world's most prestigious business graduate school is a "shitty site"? That's truly weapons-grade denialism being deployed here.

Vayra86Agentic AI for enterprise is where its at, but that is not an 'AI for most users' at all. For 'most users' its brutal overkill and totally unnecessary.

No goal-post moving allowed. "AI for most users" -- i.e. consumer-grade LLMs -- are just one single use case. Your claim was that "the vast majority of use cases" for AI are nothing but hot air. And in the '80s and even '90s I heard the same opinion expressed countless times that the Internet was total unnecessary for most users. In the '70s, the same was being said of PCs for most users.

To circle back around to the the RTX 50-series -- in the long run, consumer-grade graphics cards are benefiting dramatically from the AI revolution, even if at present it means short-term price hikes and potential shortages.

Vayra86I work for a global IT firm, I'm pretty close to the actual material in actual companies.

A predictable appeal-to-authority fallacy, but the hard data disagrees with your opinion. Every year, AI models improve dramatically not just in performance, but in speed, training efficacy, power efficiency, and any other metric one cares to name.

EndymioEvery year, AI models improve dramatically not just in performance, but in speed, training efficacy, power efficiency, and any other metric one cares to name.

Okay, but what about accuracy?
If it just throws out inaccurate/incorrect information faster - it's pointless anyway.

EndymioTo circle back around to the the RTX 50-series -- in the long run, consumer-grade graphics cards are benefiting dramatically from the AI revolution, even if at present it means short-term price hikes and potential shortages.

That sounds to me just like Blockchain evangelists, Crypto Bros saying cryptomining highs are extremely beneficial for Gaming, because of all the money GPU makers are raking in - imagine what they could develop with that, benefitting all!

And all we got was higher prices even after crypto crash, x080 cards going from $700 to $1200, and a newfound boldness that Gaming isn't really where the money's at after all...

AVATARATOkay, but what about accuracy?
If it just throws out inaccurate/incorrect information faster - it's pointless anyway.

Did you miss the word "performance" in the improvements AI models are showing each year?

BTW, when rating model performance, precision and recall are generally more important metrics than accuracy. (And no, "recall" in this context doesn't mean what you think it means).

BwazeThat sounds to me just like Blockchain evangelists, Crypto Bros saying cryptomining highs are extremely beneficial for Gaming, because of all the money GPU makers are raking in - imagine what they could develop with that, benefitting all!

And all that turned out to be accurate. If the only revenues GPU makers were receiving came from videogamers hoping to increase their headshot rate, you'd still be using a 1080ti, hoping one day a more advanced GPU might be released.

EndymioDid you miss the word "performance" in the improvements AI models are showing each year?

BTW, when rating model performance, precision and recall are generally more important metrics than accuracy. (And no, "recall" in this context doesn't mean what you think it means).

Eh, I prefer performance in my CPU/GPU, but when I'm looking for information I need accuracy so I don't do something wrong multiple times as usually happens with this AI bulshits.

AVATARATEh, I prefer performance in my CPU/GPU, but when I'm looking for information I need accuracy

When an AI model is examining your test results for cancer or screening your emails for spam you want precision more than you do accuracy. These terms have a very exact meaning in ML nomenclature, rather than their loose English definitions.

EndymioWhen an AI model is examining your test results for cancer or screening your emails for spam you want precision more than you do accuracy. These terms have a very exact meaning in ML nomenclature, rather than their loose English definitions.

Sorry, I'm just using the "replay" button, if you want the right word, just use it so when I repeat you, I'll be right.

EndymioAnd all that turned out to be accurate. If the only revenues GPU makers were receiving came from videogamers hoping to increase their headshot rate, you'd still be using a 1080ti, hoping one day a more advanced GPU might be released.

That's retarded. How did we ever got from nVidia TNT2 in 1999, borrowed money for the past three decades and hoped for crypto, AI bubbles to finally repay it?